Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection

This study develops a classification model for detecting multiple flight conditions of VTOL (Vertical Take-Off and Landing) UAVs using accelerometer data, with a motion capture system included for comparison. The objective is to identify the most effective machine learning model for classifying vari...

Full description

Bibliographic Details
Main Authors: Mohd Sani, Fareisya Zulaikha, Rohidi, Muhammad Adam, Mohd Nor, Elya, Md Ali, Syaril Azrad, Makhtar, Siti Noormiza
Format: Article
Language:English
Published: The Aeronautical and Astronautical Society of the Republic of China 2025
Online Access:http://psasir.upm.edu.my/id/eprint/117764/
http://psasir.upm.edu.my/id/eprint/117764/1/117764.pdf
_version_ 1848867334981156864
author Mohd Sani, Fareisya Zulaikha
Rohidi, Muhammad Adam
Mohd Nor, Elya
Md Ali, Syaril Azrad
Makhtar, Siti Noormiza
author_facet Mohd Sani, Fareisya Zulaikha
Rohidi, Muhammad Adam
Mohd Nor, Elya
Md Ali, Syaril Azrad
Makhtar, Siti Noormiza
author_sort Mohd Sani, Fareisya Zulaikha
building UPM Institutional Repository
collection Online Access
description This study develops a classification model for detecting multiple flight conditions of VTOL (Vertical Take-Off and Landing) UAVs using accelerometer data, with a motion capture system included for comparison. The objective is to identify the most effective machine learning model for classifying various flight conditions, such as healthy and faulty propellers, different payloads, and windy environments. Initially, various machine learning models, including Quadratic Support Vector Machine (QSVM), Neural Networks, and Naive Bayes, were trained using acceleration and displacement data. QSVM was identified as the best-performing model, achieving 87.5% training accuracy with acceleration data and 79.3% with displacement data. Following this, data from two accelerometers (an iPhone SE 2020 and an ADXL345) were used exclusively with the QSVM model for further comparison. The iPhone SE sensor achieved 97.73% training accuracy, while the ADXL345 attained 93.06%. While the iPhone sensor demonstrates superior performance, it serves only as a benchmark, as it is not intended for onboard UAV applications. The results indicate that affordable sensors, like the ADXL345, can achieve sufficient accuracy, making them viable for practical UAV deployments. The study concludes by recommending higher-quality sensors and advanced machine learning techniques for enhanced UAV fault detection.
first_indexed 2025-11-15T14:34:51Z
format Article
id upm-117764
institution Universiti Putra Malaysia
institution_category Local University
language English
last_indexed 2025-11-15T14:34:51Z
publishDate 2025
publisher The Aeronautical and Astronautical Society of the Republic of China
recordtype eprints
repository_type Digital Repository
spelling upm-1177642025-06-11T07:34:32Z http://psasir.upm.edu.my/id/eprint/117764/ Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection Mohd Sani, Fareisya Zulaikha Rohidi, Muhammad Adam Mohd Nor, Elya Md Ali, Syaril Azrad Makhtar, Siti Noormiza This study develops a classification model for detecting multiple flight conditions of VTOL (Vertical Take-Off and Landing) UAVs using accelerometer data, with a motion capture system included for comparison. The objective is to identify the most effective machine learning model for classifying various flight conditions, such as healthy and faulty propellers, different payloads, and windy environments. Initially, various machine learning models, including Quadratic Support Vector Machine (QSVM), Neural Networks, and Naive Bayes, were trained using acceleration and displacement data. QSVM was identified as the best-performing model, achieving 87.5% training accuracy with acceleration data and 79.3% with displacement data. Following this, data from two accelerometers (an iPhone SE 2020 and an ADXL345) were used exclusively with the QSVM model for further comparison. The iPhone SE sensor achieved 97.73% training accuracy, while the ADXL345 attained 93.06%. While the iPhone sensor demonstrates superior performance, it serves only as a benchmark, as it is not intended for onboard UAV applications. The results indicate that affordable sensors, like the ADXL345, can achieve sufficient accuracy, making them viable for practical UAV deployments. The study concludes by recommending higher-quality sensors and advanced machine learning techniques for enhanced UAV fault detection. The Aeronautical and Astronautical Society of the Republic of China 2025 Article PeerReviewed text en http://psasir.upm.edu.my/id/eprint/117764/1/117764.pdf Mohd Sani, Fareisya Zulaikha and Rohidi, Muhammad Adam and Mohd Nor, Elya and Md Ali, Syaril Azrad and Makhtar, Siti Noormiza (2025) Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection. Journal of Aeronautics, Astronautics and Aviation, 57 (3). pp. 737-749. ISSN 1990-7710 https://www.airitilibrary.com/Article/Detail/P20140627004-N202504100011-00045 10.6125/JoAAA.202503_57(3S).44
spellingShingle Mohd Sani, Fareisya Zulaikha
Rohidi, Muhammad Adam
Mohd Nor, Elya
Md Ali, Syaril Azrad
Makhtar, Siti Noormiza
Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection
title Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection
title_full Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection
title_fullStr Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection
title_full_unstemmed Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection
title_short Comparison of classification models and accelerometer sensors for VTOL UAV flight condition detection
title_sort comparison of classification models and accelerometer sensors for vtol uav flight condition detection
url http://psasir.upm.edu.my/id/eprint/117764/
http://psasir.upm.edu.my/id/eprint/117764/
http://psasir.upm.edu.my/id/eprint/117764/
http://psasir.upm.edu.my/id/eprint/117764/1/117764.pdf